Wikipedia talk:Suspend use of stub icons

Suspend most stub icons

I agree than stub icons put a strain on servers, and for that reason, WP:WSS has reduced the use of icons on many large stub categories. Some icons, however, serve a definite purpose. A clear example is those on the five regional geography subs for African regions, and the geography stub for central america. The maps on these items clearly delimit what countris are meant to have those stub templates applied, and to a very great extent they work: Mexican subs, which might otherwise find themselves added to the Central American category, do not find themselves in that category; stubs relating to countries close to the line demarking the boundary between West and North Africa are placed in the correct category almost every time. In these cases the icons are a very useful addition to the stub templates. Grutness...wha? 05:45, 19 May 2005 (UTC)[reply]

No, all stub icons

[edit]

Sure geographic stub icons are useful. All stub icons are useful, if only to make it very clear that an article is incomplete. That's not the point. It isn't a question of useful-versus-not-useful. It's a question of whether the usefullness outweighs the anti-usefulness of not being able to browse stub categories -- which I often can't! ¶ I do agree that we need to hear from a developer before we proceed. ----Isaac R 22:31, 19 May 2005 (UTC)[reply]

It's perhaps worth noting that the ones I named are relatively small categories - none of them have over 400 articles, and some are very close to the point where subcategories may be split off them. In most cases, the icons are largely irrelevant, as it is the templates that remind us that the article is incomplete. And a stub template without an icon is still a stub template. There are, however, a very few templates (maybe 5%), like the ones I mentioned, where additional important information is given by the icon. Unless - of course, you'd like to see a template that reads: This article on a location in Central America, i.e., somwhere in Belize, Honduras, Nicaragua, Guatemala, Costa Rica, El Salvador, or Panama (but not Mexico or the Caribbean), is a stub. Grutness...wha? 01:31, 20 May 2005 (UTC)[reply]

Developer's opinion

[edit]

Can we get a developer's opinion here please? It seems like a good idea in principle. R adiant _* 07:39, May 19, 2005 (UTC)

Very true. It seems that everyone's an armchair server admin these days. I'm all for removing unsreasonable server strain, but I'd like to have this backed up with some numbers by someone who's actually a developer, because I'm not convinced that the stub images are that much more horrible than everything else. "I think I remember Jamesday mentioning it once on IRC about 6 months ago" doesn't count, BTW. -- grm_wnr Esc 08:10, 19 May 2005 (UTC)[reply]

Fair enough. Come to think of it, it does seen to have been handed down in Wiki lore that icons are a strain... about time we had some definitive info. Grutness...wha? 09:59, 19 May 2005 (UTC)[reply]

Do any of the three of you know where to find a developer with the requisite knowledge to address this? Courtland 22:12, 2005 May 19 (UTC)

I'd recommend either wikitech-l or #mediawiki. I've spent most of yesterday trying to get an admin to help me, so I'm leaving this one to others ;) -- grm_wnr Esc 23:29, 19 May 2005 (UTC)[reply]

For some info about images in templates, see m:Image server overload 2005-03 and some of the discussion at Wikipedia talk:Avoid using meta-templates by the main WikiMedia developer Jamesday. BlankVerse ∅ 13:51, 20 May 2005 (UTC)[reply]

Thanks. I think what is relevant here is the following post by Jamesday:

Can't count right now. Too many for comfort. Images in templates are a significant performance issue at present. Work to change the software to try to make them less of an issue is happening. Have been two tries so far because the image server load was making the site comletely unavailable at times. Temporarily removing the images from the most popular templates helped quite a bit. See [1]. Too soon to say whether the second software approach will help enough - buying another, more powerful, server anyway, I think, because it's a critical problem we can't allow to continue. It might end up leading to us turning off image display completely just to keep the sites available if we don't get it dealt with. Temporarily removing the images from the most popular stubs did enough good that we haven't yet been forced to do that. Hopefully either the software changes will do enough good or the new hardare will do enough good before we have to do something more dramatic. Longer term more software changes are coming (distributing the images over many servers). Jamesday 15:24, 7 Apr 2005 (UTC)

They are still pretty problematic right now, but that this may be alleviated with the next MediaWiki upgrade - and 1.5 is scheduled for June 1, with its flagship installation en.wikipedia probably upgraded to that as soon as possible. So I propose waiting until then before we make a policy. -- grm_wnr Esc 15:54, 20 May 2005 (UTC)[reply]

I couldn't find it on a quick search, but somewhere Jamesday also talked about new hardware for the graphics server. I'd say that just as a precautionary measure, icons should be removed from any stub template that is on > 500 articles. BlankVerse ∅ 16:17, 20 May 2005 (UTC)[reply]

Out of curiosity (because I didn't understand the technical issue completely), I just asked developer User:Tim Starling on IRC about the performance hit of images in templates - he said that the problems are not images in templates, but images in general, because the apache server has to load the image header from the image server to get the images' dimensions (which can be cached but the cache gets cleared sometimes). This is fixed in 1.5, in which the information is put in the database on upload. And I quote: "because of the cache, commonly used images shouldn't be a problem anyway - the request to remove images from templates came from before we had the cache". -- grm_wnr Esc 16:26, 20 May 2005 (UTC)[reply]

Our image server (albert) has been in pretty good shape since we implemented the image metadata cache. Haven't you noticed that images are usually fast to load these days, even in peak time? Whenever anything goes wrong with memcached, albert is the first thing to die, but that's a rare event.

I think template images are being unfairly targetted here. Nobody (Jamesday included) has studied the amount of traffic created by template images. I would think that the most commonly requested images would be photos on popular articles such as the featured article, rather than icons on obscure categories. I think these icons make an important contribution to Wikipedia's visual appeal, and I would prefer to support their inclusion by making the necessary software changes, rather than try to enforce their removal. -- Tim Starling 16:44, May 20, 2005 (UTC)

This isn't about fairness, this is about making the system responsive. I didn't start this RFC because I have some anti-icon agenda. I actually consider them very useful and I'd love to see them more widely used. The only question is, how high a price are we paying to use them? ----Isaac R 20:13, 20 May 2005 (UTC)[reply]

How about a proposal for an experiment. An image could be added to one or more large categories (grins) {{stub}} itself might be a good test at a defined time for a defined period to see what kind of performance hit is experienced. The tally for {{stub}} is ~4000 articles (20 pages); the experiment could be scaled up into the 10,000 range probably as well by adding in some of the other large types. Does this sound like a useful exercise, or am I just barking up the WP:POINT tree :), albeit for the good of Wiki-kind? Courtland 01:04, 2005 May 21 (UTC)

Oh, I just thought of something else ... it is not unlikely that the stub sorters themselves create the greatest number of stub-article page calls. If one were to say that during a time overlapping the experiment period all the stub sorters should concentrate their efforts on {{stub}} you'd have a pretty hefty load test I think. Courtland 01:07, 2005 May 21 (UTC)

I'm not sure that I want to get into the habit of recommending "experiments" with the Wikipedia Database, but for gauging the impact of images in templates, it would probably be best to look at a small number of VERY popular templates (the Sister Templates?), and look at database performace with and without the images. I wonder if there is any sort of list of the most frequently used templates? (Also, I want to express a thanks to developer Tim Starling for helping to explain things.) BlankVerse ∅ 04:10, 21 May 2005 (UTC)[reply]

The Story So Far

[edit]

The word from developers seems to be that, yes, template icons do impact performance -- but they're hoping to fix this. At this point I'm all for waiting to see if they can. But if it turns out they can't, I'm going to renew my original proposal. Which might be better phrased this way: eliminate all stub icons, not just "popular" or "not useful" ones, because it's too hard to identify which ones are popular or useful. Plus it creates bad feelings by users who see their favorite icons eliminated, while others continue to use them.

On the other hand, if it can be established that we can use template icons without bringing the image servers to their knees, I'd want to see an initiative to put icons in all stub templates. These icons are very useful not just for categorizing stubs, but for making casual readers aware that articles are incomplete. ----Isaac R 18:55, 3 Jun 2005 (UTC)

Um, if you read what Tim wrote above you'll see that this is pretty much a non-issue. Further, two new high-power db servers are being installed within days; as a cascade effect I expect the image server will be boosted. Dan100 21:39, Jun 9, 2005 (UTC)

Abandoned?

[edit]

I placed the rejected template as I got the impression from the above that this is abandoned for the time being, until further details clarify the situation. Also removed from RfC Hiding 21:36, 20 Jun 2005 (UTC)

Status?

[edit]

Has anything been decided? Has new information presented itself? Has it been decided that it is a good idea to remove images from stub templates?

It appears that for now, the issue has been resolved with the cache solution, and that the strain images put on the server has been greatly reduced; I would prefer it if a developer could add facts to this impression of mine.

ᓇᐃᑦᔅᑕᓕᐅᓐ ✉ 13:10, 29 December 2005 (UTC)[reply]